|Home | About | Journals | Submit | Contact Us | Français|
Contributors: CV and ST-A carried out the research on which the original protocol was based. All authors participated equally in the development of the protocol, in which successive versions were tested in clinical practice and refined in the light of experience. The writing of the original protocol and present paper was primarily carried out by CV, ST-A, EJC, and DH, but all authors contributed to the final version. CV and DH are the guarantors.
Why do things go wrong? Human error is routinely blamed for disasters in the air, on the railways, in complex surgery, and in health care generally. However, quick judgments and routine assignment of blame obscure a more complex truth. The identification of an obvious departure from good practice is usually only the first step of an investigation. Although a particular action or omission may be the immediate cause of an incident, closer analysis usually reveals a series of events and departures from safe practice, each influenced by the working environment and the wider organisational context. This more complex picture is gaining acceptance in health care,1,2 but it is seldom put into practice in the investigation of actual incidents.
The Clinical Risk Unit has developed a process of investigation and analysis of adverse events for use by researchers.3–7 Two years ago a collaborative research group was formed between the unit and members of the Association of Litigation and Risk Management (ALARM). This group has adapted the research methods to produce a protocol for the investigation and analysis of serious incidents for use by risk managers and others trained in incident analysis. The protocol gives a detailed account of the theoretical background and process of investigation and analysis together with case examples.8 In this article we introduce the main ideas and present a section of a case analysis to illustrate the methods used.
The theory underlying the protocol and its application derives from research in settings outside health care. In the aviation, oil, and nuclear industries, for instance, the formal investigation of incidents is well established.9 Studies in these areas and in medicine have led to a much broader understanding of the causes of accidents, with less focus on the individual who makes the error and more on pre-existing organisational factors. Such studies have also illustrated the complexity of the chain of events that may lead to an adverse outcome.4,6,10–12 The root causes of adverse clinical events may lie in factors such as the use of locum doctors and agency nurses, communication and supervision problems, excessive workload, and educational and training deficiencies.
In health care the development of prevention strategies from such analyses has not yet been fully exploited. However, the potential for these approaches is apparent in other domains. For instance, the inquiry into the Piper Alpha oil disaster led to a host of recommendations and the implementation of several risk reduction strategies throughout the industry. These included the setting up of a single regulatory body for offshore safety, relocation of pipeline emergency shutdown valves, the provision of temporary safe refuges for oil workers, new evacuation procedures, and requirements for emergency safety training. Most interestingly, oil companies were required to demonstrate that hazards had been minimised and were as low as could reasonably be expected.13–15
In the original research we used a combination of record review, staff interviews, and a checklist of human factors highlighting psychological and organisational factors. We applied Reason's model of organisational accidents to clinical incidents,2 reviewing both errors made and the background organisational factors that were implicated. We have extended Reason's generic model and adapted it for use in a healthcare setting, classifying the error producing conditions and organisational factors in a single broad framework of factors affecting clinical practice. Both Reason's model and our framework are described elsewhere.5 The figure and table table11 show the essential components of the model and the framework.
The model requires that the investigator starts by examining the chain of events that led to an accident or adverse outcome and considering the actions of those involved. The investigator then, crucially, looks further back at the conditions in which staff were working and the organisational context in which the incident occurred. Below, we detail the application of the model in a healthcare setting.
The first step in any analysis is to identify the care management problems, which are broadly speaking the healthcare equivalent of Reason's “unsafe acts.”2,7 Care management problems are actions or omissions by staff in the process of care. These may be slips, such as picking up the wrong syringe, lapses of judgment, forgetting to carry out a procedure, or, rarely, deliberate departures from safe operating practices, procedures, or standards.
Care management problems have two essential features: firstly, care deviated beyond safe limits of practice and, secondly, the deviation had a direct or indirect effect on the eventual adverse outcome for the patient. (In cases where the impact on the patient is unclear it is sufficient that the care management problem had a potentially adverse effect.) Each care management problem is identified individually and the reason for its occurrence analysed separately.
Failure to monitor, observe, or act
Delay in diagnosis
Incorrect risk assessment (for example, of suicide or self harm)
Failure to note faulty equipment
Failure to carry out preoperative checks
Not following an agreed protocol (without clinical justification)
Not seeking help when necessary
Failure to supervise adequately a junior member of staff
Incorrect protocol applied
Treatment given to incorrect body site
Wrong treatment given
For each care management problem identified, the investigator records the salient clinical events or condition of the patient at that time (for example, bleeding heavily, blood pressure falling) and other patient factors affecting the process of care (for example, patient very distressed, patient unable to understand instructions).
All investigations consist of a series of steps that should be followed, as a matter of routine, when an incident is investigated:
1. Ascertain that a serious clinical incident has occurred and ensure it is reported formally. Alternatively identify an incident as being fruitful in terms of organisational learning
2. Trigger the investigation procedure. Notify senior members of staff who have been trained to carry out investigations
3. Establish the circumstances as they initially appear and complete an initial summary, decide which part of the process of care requires investigation, prepare an outline chronology of events, and identify any obvious care management problems
4. Structured interview of staff:Establish chronology of eventsRevisit sequence of events and ask questions about each care management problem identified at the initial stageUse framework to ask supplementary questions about reasons for each care management problem
5. If new care management problems have emerged during interviews add them to initial list. Interview again if necessary
6. Collate interviews and assemble composite analysis under each care management problem identified. Identify both specific and, where appropriate, general contributory factors
7. Compile report of events, listing causes of care management problems and recommendations to prevent recurrence
8. Submit report to senior clinicians and management according to local arrangements
9. Implement actions arising from report and monitor progress
After the care management problems have been identified, the investigator considers the conditions in which errors occur and the wider organisational context. Using the framework (table (table1)1) both during staff interviews and afterwards, the investigator identifies the factors that led to each care management problem. Any combination of different level factors might contribute to the occurrence of a single care management problem.
Once the investigator has identified the various factors that contributed to the incident, a further distinction needs to be drawn between specific contributory and general contributory factors. For instance, a failure of communication between a doctor and a nurse may have contributed to a care management problem. If this is an isolated occurrence, then it is a specific contributory factor with no more general implications. If, on the other hand, the problem is quite common the investigator would also note a general contributory factor of “poor communication.” The general factors are those longstanding features which have clear implications for the safe and effective running of the unit or hospital. When trying to decide whether to record a general contributory factor the investigator might consider such questions as:
More details on the investigation process are available on the BMJ 's website and in the protocol.8 The box below provides a summary.
The box on the next page shows an example based on real clinical events, but it has been altered in various respects to preserve anonymity. The case is described and fully analysed in the protocol.8 The main care management problems identified were non-communication of formulated care plan, inadequate fetal monitoring in first and second stage of labour, inadequate pain control in first stage of labour, and delay in management in second stage of labour. Each of these care management problems was analysed separately. Table Table22 shows the contributory factors for the problem with fetal monitoring.
The final report of this case concluded that “with the benefit of hindsight, the outcome of this delivery might, on a balance of probabilities, have been different.”8 After analysis of this case and discussion of the implications, changes were made to the organisation and policies of the unit. These included a new protocol stipulating that when there was a conflict between information provided by different types of monitoring equipment, best practice would be to assume the worst case and seek medical advice; an individual training programme for specific members of staff; a programme of further education for all midwives in the assessment and management of shoulder dystocia; and review and eventual replacement of all outdated fetal monitoring equipment.
Mrs B was booked for shared care. Her last child weighed 4.4 kg at birth and slight shoulder dystocia was noted at delivery. Mrs B was referred to the consultant by the community midwife at 38 weeks as the baby felt large for dates. The ultrasound scan estimated the weight of the baby as 4.5 kg. A graded response to the findings on palpation and ultrasound was made bearing in mind the woman's previous obstetric history. Firstly, the pregnancy should not progress more than six days beyond the due date before induction of labour, rather than the usual 12-14 days. Secondly, it was recorded that no attempt should be made at a difficult mid-cavity instrumental delivery. Thirdly, the possibility of shoulder dystocia was anticipated and recorded to forewarn the labour ward staff.
0555: Mrs B was admitted with ruptured membranes. Labour started shortly afterwards
0650: Vaginal examination showed her cervix to be 3 cm dilated. The fetal heart was monitored by external Doppler probe. At this stage Mrs B requested an epidural, but the anaesthetist was not immediately available as he was finishing handing over on the intensive care unit. Mrs B's labour proceeded rapidly and therefore an epidural was not carried out
0715: A scalp electrode was placed on the baby's head as the midwives were unable to monitor the fetal heart easily in view of maternal size and maternal distress. The trace showed the fetal heart rate to be normal
0750: A further vaginal examination was carried out. Mrs B's cervix was 6 cm dilated, the fetal heart rate was normal with good variability. Pethidine was administered
0805: The cervix was fully dilated and pushing started. Mrs B was unable to cooperate with staff as she was in pain and very distressed
0814: Scalp electrode was removed as the head was crowning. The final readings of the fetal heart before the scalp electrode was removed showed marked decelerations with a decreasing trend. The delivery did not proceed and the head remained stationary. The external Doppler probe was reattached and showed fetal heart rate at 160-170 beat/min
0833: Medical help was sought. The obstetric registrar and the duty consultant came immediately and quickly diagnosed shoulder dystocia. They carried out a McRoberts manoeuvre and then applied suprapubic pressure; the baby was delivered at 0839
0839: The infant was severely compromised with no heart beat. He was resuscitated and ventilated and then transferred to the special care baby unit but died the next day
The method described above has been tested on over 40 incidents, initially in a research context and later by clinicians and risk managers. Incidents have been investigated in obstetrics, anaesthetics, accident and emergency, orthopaedics, general medicine, and psychiatry. A structured and systematic approach means that the ground to be covered in any investigation is already largely mapped out. The protocol helps to ensure a comprehensive investigation and facilitates the production of formal reports. Although the process may initially seem complicated and time consuming, our experience is that using the protocol actually speeds up complex investigations by focusing the investigators on the key issues and bringing out the systemic factors behind the incident. Members of the research team have found that once the general contributory factors are identified, they lead automatically to the implications and action points. The final report “almost writes itself.”
Even experienced clinicians find that following the protocol brings additional benefits in terms of comprehensiveness and investigation expertise. Clinicians are accustomed to identifying the problematic features in the management of a case and so can easily identify the care management problems. However, they are less familiar with identifying contributory factors for each care management problem. A systematic approach pays dividends when exploring these. The protocol does not attempt to supplant clinical expertise. Rather, the aim is to use clinical experience and expertise to the fullest extent.
A formal, systematic approach also benefits the staff involved. The methods used are designed to promote a greater climate of openness and to move away from finger pointing and the routine assignment of blame. This differs greatly from the quasijudicial approach often used in formal inquiries. If a consistent approach to investigation is used, staff tend to find the process less threatening than traditional unstructured approaches, especially when the same procedure is being followed with everyone involved.
Early experience with the protocol has shown that some formal training and practice is needed for it to be fully effective. Although the basic ideas can be grasped relatively quickly, the full method takes longer to absorb. Guided practice on the investigation of incidents, preferably in a local context, is essential to become familiar with the methods. Initially this protocol is likely to be used by risk managers, with additional clinical input. However, we suggest that the next step is to designate and train investigators in each clinical area who can carry out an investigation to agreed guidelines.
Although we believe that the protocol is an effective and valuable tool, it is still at a relatively early stage of development, both conceptually and practically. Formal evaluation is needed and a great deal more practical testing is required. We plan to revise and develop the protocol in the light of experience and formal evaluation. The protocol also has potential as a research instrument since use of a common method will make analyses of case series of incidents more powerful. In the meantime, however, it is already proving a powerful means of investigating and analysing clinical incidents and drawing out the lessons for enhancing patient safety.
Copies of the full protocol and details of training programmes are available from Association of Litigation and Risk Management (ALARM), Royal Society of Medicine, 1 Wimpole Street, London W1.
Competing interests: CV received funding from Healthcare Risk Resources International to support the work of ST-A during the development of the protocol.
website extra: Further details of the investigation process are available on the BMJ's website www.bmj.com